Hi all,

I've a question about the acoustic waves which occurs during charge exchange in an air intake manifold. Could someone tell me if it is possible to simulate and visualize the acoustic waves in 3D as a function of crank angle?

thx

Check out web sites such as

http://www.erc.wisc.edu/

The state of the art in engine modeling is to solve the Navier-Stokes equations in the full time-dependent geometry of the engine. This will automatically include acoustic waves in the solution.

You need to have a 3D compressible Navier Stokes solver in order to propagate the acoustic pressure waves. Sound speed is infinite in an incompressible approach. This will be important in Air Intake type applications since there will be cavity modes. We have been doing such simulations a few years ago with reasonnable results.

The difficulty will arise if the throttle valve is almost closed because there will be a conflict between a super fine modeling needed around this part and the long time constant of the air box plenum. In this case, we suggest to split the problem in two parts.

Check our website www.mcube.fr in the M-Explicit section for examples and contact me directly if you have further questions

Dimitri Nicolopoulos dimitri@mcube.fr

Thank you for yours fast reply. I know that I have to use a 3D compressible solver but the problem is that all commercial cfd codes only have second order precision in discretisation. This is by far not enough for acoustic waves. I guess 7th or 8th order would be neccessary to see the acoustic pressure fluctuation.

I guess I have to couple Fluent with sysnoise or comet acoustics to obtain a possible solution but I'am not sure if I'll see the wave reflextions.

rgds

It is not possible to calculate sound waves directly by solving the NS equation even if you use a DNS because the amplitude of the these waves is at least an order of magnitude lower than the turbulent pressure fluctuations (also known as "speudo sound") and below the round-off error of your machine.

It should be also interesting for you to read the following article about higher order approximation:

http://www.cfd-online.com/Forum/main...cgi?read=10138

Yes thank you.If have always asked myself why there isn't a tool like wave or gt-power for 3D. Now it is clear.

thx

I kind of disagree with some of the posts in this thread:

1)I dont think that 7th or 8th order of precision of the numerical scheme is needed. Actually, a right combination of numerical scheme order and element size is needed. However, we believe that using 64 bits arithmetics is critical to be able to capture properly both the flow and acoustic waves. The second critical ingredient in our opinion is to use a small time step (microsec or so) to make sure the propagations are captured properly.

2)CAA is a challenging task, requiring lots of CPU and detailed attention to build good numerical models. However, I would not rule it out saying it's impossible! For those who are interested in seeing simulations compared to experiments for internal and external CAA problems, download the very detailed paper we wrote about CAA and the way we see it. You will see examples coming from our clients, namely PSA and HONDA:

http://www.mcube.fr/M-Cube/post/CFD-Online.pdf

I'm looking forward to answer the questions you might have about our experience.

Regards,

Dimitri

Yes, it is possible to compute sound waves in a NS solver, but it is difficult and requires special numerical schemes and boundary conditions. You can download a couple of good papers on LES computations of jet noise from here:

http://www.tfd.chalmers.se/~lada/pro.../proright.html

For the vast majority of applications though using NS to compute acoustic waves is overkill.

First of all thank you for your precise explanations. But I've still a qustion about this topic. In our company we have star-cd and fluent as cfd solvers. What actually will I receive when fluent or star-cd is coupled with for example the newest version of sysnoise. I think sysnoise can read in transient cfd data (LES t.m. must used). Actually the acoustic waves as a function of time should be calculated but now I am in doubt.Could someone tell me if this is a possible result?

thx

Using Fluent or Star + Sysnoise is a workaround. If you try to go beyond 1-200 Hz it will most likely not work and if you stay incompressible you will not be able to identify cavity modes.

In other words, there are no free lunches... mixing a CFD code + an Acoustic code does not produce an Aero-Acoustic solver!

Once again, you should download our white paper on our website www.mcube.fr which attempts to discuss the ins and the outs of CAA simulations.

Regards,

Dimitri